taichi.linalg#
- class taichi.linalg.SparseMatrix(n=None, m=None, sm=None, dtype=f32)#
Taichi’s Sparse Matrix class
A sparse matrix allows the programmer to solve a large linear system.
- Parameters
n (int) – the first dimension of a sparse matrix.
m (int) – the second dimension of a sparse matrix.
sm (SparseMatrix) – another sparse matrix that will be built from.
- shape(self)#
The shape of the sparse matrix.
- transpose(self)#
Sparse Matrix transpose.
- Returns
The transposed sparse mastrix.
- class taichi.linalg.SparseMatrixBuilder(num_rows=None, num_cols=None, max_num_triplets=0, dtype=f32)#
A python wrap around sparse matrix builder.
Use this builder to fill the sparse matrix.
- Parameters
num_rows (int) – the first dimension of a sparse matrix.
num_cols (int) – the second dimension of a sparse matrix.
max_num_triplets (int) – the maximum number of triplets.
- build(self, dtype=f32, _format='CSR')#
Create a sparse matrix using the triplets
- print_triplets(self)#
Print the triplets stored in the builder
- class taichi.linalg.SparseSolver(dtype=f32, solver_type='LLT', ordering='AMD')#
Sparse linear system solver
Use this class to solve linear systems represented by sparse matrices.
- Parameters
solver_type (str) – The factorization type.
ordering (str) – The method for matrices re-ordering.
- analyze_pattern(self, sparse_matrix)#
Reorder the nonzero elements of the matrix, such that the factorization step creates less fill-in.
- Parameters
sparse_matrix (SparseMatrix) – The sparse matrix to be analyzed.
- compute(self, sparse_matrix)#
This method is equivalent to calling both analyze_pattern and then factorize.
- Parameters
sparse_matrix (SparseMatrix) – The sparse matrix to be computed.
- factorize(self, sparse_matrix)#
Do the factorization step
- Parameters
sparse_matrix (SparseMatrix) – The sparse matrix to be factorized.
- info(self)#
Check if the linear systems are solved successfully.
- Returns
True if the solving process succeeded, False otherwise.
- Return type
bool
- solve(self, b)#
Computes the solution of the linear systems. :param b: The right-hand side of the linear systems. :type b: numpy.array or Field
- Returns
The solution of linear systems.
- Return type
numpy.array
- class taichi.linalg.sparse_matrix_builder#